• Fixed a bug in the generated JSP code for the def opcode in ll.xist.ns.jsp.fromul4.

• Fixed two bugs in the XML codecs:

  • An externally specified encoding wasn't honored in the incremental decoder.

  • Fixed reset() for incremental codecs: If encoding has been changed during parsing in the incremental codecs it now gets reset to its proper initial value.

• Fixed a bug in the handling of the UL4 opcode addlist in ll.xist.ns.jsp.fromul4.

• Added missing processing instruction class for the UL4 def tag to the ll.xist.ns.ul4 namespace module.

• The generated JSP code for the loadvar opcode now uses the Java method Utils.getItem, so that non-existent variables no longer get treated as None/null.

• ll.xist.ns.jsp.fromul4 now outputs the correct code for calling the format method on date objects (This requires version exp-22 of the UL4 Java package).

• In UL4 templates it's now possible to define locale templates via <?def tmpl?>templatecode<?end def?>.

• Python 2.6 is required now.

• ll.orasql and ll.nightshade are now part of the distribution.

• ll.make has a new Action class: ObjectAction simply returns an existing object.

• The following classes have been removed from ll.make: EncodeAction, DecodeAction, EvalAction, GZipAction, GUnzipAction, JavascriptMinifyAction, XISTBytesAction, XISTStringAction, JoinAction, UnpickleAction, PickleAction, TOXICAction, TOXICPrettifyAction, SplatAction, UL4CompileAction, UL4RenderAction, UL4DumpAction, UL4LoadAction, XISTTextAction and XISTConvertAction. All of these actions can be executed by using CallAction or CallAttrAction.

• ll.make.PipeAction has been renamed to TransformAction.

• The new ll.make.PipeAction pipes the input through an external command.

• ll.make.FileAction now automatically wraps the key argument into an URL object.

• ll.make.FileAction has two new methods chmod and chown that return a ModeAction and OwnerAction for modifying the file created by the FileAction.

• ll.make.Action has three new methods: call, getattr and callattr create a CallAction, GetAttrAction or CallAttrAction object respectively.

• The division operator is no longer implemented for Action objects in ll.make.

• Two new UL4 functions have been added: float and iscolor.

• Two new scripts have been added: uls can be used to list any directory given as an URL. ucat can be used to output any file or directory.

• The script ucp now changes the user and group only if a user or group is given.

• A bug in the 64-bit support for sgmlop has been fixed.

• Fixed a bug in the remote stat method for ssh URLs (it seems that the posix.stat_result tuple objects can no longer be pickled).

• There's a new function misc.itersplitat for splitting a string at specified positions.

• For ssh URLs a keyword argument ssh_config is supported now instead of identity (This mirrors the corresponding change in the py library)

• Fixed handling of empty pid files in ll.sisyphus (Fixes issue #11 reported by Jarek Zgoda).

• Fix UL4 templates that produce no output: As the generated Python sourcecode didn't contain any yield statements, the resulting function was an ordinary function instead of a generator.

• A new UL4 method join has been added. It works like the Python string method join.

• ll.misc has three new functions: gzip und gunzip can be used for compressing and uncompressing byte strings with gzip. jsmin can be used to minify Javascript source.

• Parsing an empty string with tidy=True in ll.xist.parsers.parsestring now works again.

• The xfind operators attrhasvalue, attrhasvalue_xml, attrcontains, attrcontains_xml, attrstartswith, attrstartswith_xml, attrendswith, attrendswith_xml, hasid and hasclass now support multiple values. The operator matches the node if it matches with any of the given values.

• A new function reversed is now available in UL4 templates. It returns an iterator that will output the items of any sequence in reverse order.

• Inside UL4 templates rendering other templates can now be done with the new render method. This method returns the template output as a string. Passing parameters can be done via keyword arguments or with the ** syntax like when using the render tag.

• A new version of the int function has been added to UL4: When called with two arguments, the first must be a string, and the second is treated as the base for the conversion.

• Generating the Python source from an UL4 template is now 20-25% faster.

• Fixed a buffer overrun in the C portions of the url module.

• Added a class addattr to ll.xist.xsc. This can be used to extend XML attributes via with blocks.

• Added the function ll.xist.ns.jsp.fromul4 which can turn an UL4 template into JSP source code.

• The following Color class methods have been dropped: fromrgba, fromrgba4, fromrgba8, fromint4, fromint8.

• The following Color properties have been dropped: r4, g4, b4, a4, r8, g8, b8, a8, r, g, b, a int4, int8, rgb4, rgba4, rgb8, and rgba8 have been dropped. The new methods r, g, b and a return the 8 bit component values.

• The class methods fromhsva and fromhlsa have been renamed to fromhsv and fromhls.

• The property css has been dropped. Instead the CSS string is returned by __str__.

• Dividing color now does a scalar division. Blending colors is now done with the modulo operator.

• Support for color objects has been added to UL4.

• The XPIT templating language and ll.make.XPITAction have been removed.

• Fixed a bug in ll.make.CacheAction.get: The action must return real data when called with bigbang as the timestamp.

• ll.make.UL4RenderAction has been fixed.

• A new function json has been added to UL4: This function returns a JSON dump of the object passed in (this requires either simplejson or Python 2.6).

• The UL4 function csvescape has been renamed to csv.

• A new option --showregistration/-r has been added to make scripts.

• ll.make now supports Growl notifications on Mac OS X. To activate it set the LL_MAKE_GROWL environment variable to 1 or use the -g or --growl options.

• ll.make has a new action class JavascriptMinifyAction for minimizing Javascript source.

• ll.color.Color has been rewritten to create immutable objects with the components being 8 bit values (i.e. 0-255) instead of floating point values between 0 and 1. An alpha component has been added.

• A strong element has been added to the ll.xist.ns.doc namespace.

• Fixed a bug in ll.make.JoinAction.execute.

• css.applystylesheets could no longer handle style declarations containing comments. This has been fixed now.

• Parsing didn't work when tidy was set to true and a base argument was given. This has been fixed now.

• Bugs with thread local storage have been fixed so using xsc.Pool, xsc.build and URL contexts in with blocks in multithreaded applications should work now.

• Templates can no longer be passed as a separate dictionary to UL4 templates but are passed as variables like other data objects too.

• Strings in UL4 have gained a new method capitalize.

• Printing XML escaped strings in UL4 has now gained its own tag and opcode. <?printx foo?> is equivalent to <?print xmlescape(foo)?>.

• Exception handling in UL4 has been rewritten to allow proper error reporting when calling nested templates.

• UL4 has gained a new function zip. It can be called with two or three arguments and does what itertools.zip does.

• UL4 has gained another new function: type returns the type of its argument as a string.

• UL4 now supports tuple unpacking in <?for?> tags with three variables.

• UL4 has a new tag for comments: <?note This is comment?>.

• A new script db2ul4.py has been added that can render UL4 templates with database content.

• In UL4s <?render?> tags it's now possible to pass along a complete argument dictionary via the **arg syntax just like in Python. This syntax can even be used multiple times in the call. This syntax is available in dictionary literals too, i.e. {1:2, 3:4} and {**{1:2}, **{3:4}} are equivalent.

• A new UL4 function get has been added that works similar to the dictionary method get, but works with global variables.

• The missing processing instruction render has been added to ll.xist.ns.ul4.

• xml_codec now partially works, even if the C module is missing. As long as you explicitly specify an encoding on parsing and publishing it should work.

• A new processing instruction class ll.xist.AttrProcInst has been introduced. When an AttrProcInst node is the first node in an attribute, it takes over publishing of the attribute. In all other cases the processing instruction disappears completely. UL4 uses this to implement "conditional attributes" (via the new classes attr_if and attr_ifnn).

• Building trees with with blocks has changed slightly. Nodes used in with blocks and with + are now passed to a with handler instead of building the tree directly. This fixes a problem when nested convert calls use with blocks.

• The element ll.xist.ns.form.memo has been renamed to textarea and ll.xist.ns.form.edit has been renamed to text. Classes ll.xist.ns.form.button and ll.xist.ns.form.file have been added.

• Iterating through the inputs in ll.make actions has been fixed (i.e. the additional inputs will be output too). ll.make.Project.findpaths has been fixed to work with non-ll.make.Action inputs. (This means that now you have to pass a real registered target action to findpaths not just its key).

• ll.make has gained a new action: XISTStringAction publishes an XIST node as a unicode string. XISTPublishAction has been renamed to XISTBytesAction.

• Fixed a bug in the caching logic in ll.make.CacheAction.

• ll.make.CallMethAction has been renamed to CallAttrAction because it can be used to e.g. call functions in a module too.

• The properties showaction, showstep and shownote of ll.make.Project object can now be assigned booleans as well (which results in all or no actions being shown.

• The version number for cssutils has been bumped to 0.9.5.1.

• Dictionaries in UL4 have gained a new method get.

• The version number for cssutils has been bumped again (to 0.9.5rc2 or a later 0.9.5 version).

• Fixed a bug in the parsing of slice expressions in UL4.

• ll.make has gained a new UL4RenderAction action.

• Fixed a bug in the formatting for the getslice2 opcode for UL4.

• Fixed a bug in the implementation of the "not" operator in UL4.

• When the UL4 compiler encounters unclosed blocks, it will now include the start location of the block in the error message.

• XIST has gained its fourth templating language: UL4 the "Universal Layout Language". This templating language is similar in capabilities to Djangos templating language. However UL4 templates are compiled to a bytecode format, which makes it possible to implement template renderers in other languages and makes the template code "secure" (i.e.template code can't open or delete files).

• ll.make has gained new actions: GZipAction, GUnzipAction, CallFuncAction, CallMethAction, UL4CompileAction, UL4DumpAction and UL4LoadAction.

• The version number for cssutils has been bumped to 0.9.5rc1.

• Nodes of type ll.xist.xsc.Comment and ll.xist.xsc.DocType inside of attributes are now simply ignored when publishing instead of generating an exception.

• All actions in ll.make no longer check whether their inputs are action objects. Non-action objects are simply treated as ancient input data. This also means that most action classes have an input parameter in their constructor again, as this input could now be a constant.

• Most attributes of action objects in ll.make can now be action objects themselves, so for example the name of the encoding to be used in an EncodeAction can be the output of another action.

• ll.make.ImportAction has been dropped as now the module object can be used directly (e.g. as the input for an XISTPoolAction object).

• ll.misc.xmlescape now escapes ' as ' for IE compatibility.

• Functions ll.misc.xmlescape_text and ll.misc.xmlescape_attr have been added that implement the functionality from XIST 3.2.5 and earlier.

• The default parser for XIST is expat now. To switch back to sgmlop simply pass an SGMLOPParser object to the parsing functions:

  >>> from ll.xist import parsers
  >>> node = parsers.parsestring("<a>", parser=parsers.SGMLOPParser())

• TOXIC has been split into a compiler module ll.toxicc and an XIST namespace ll.xist.ns.toxic. TOXIC now supports output for SQL Server. The function xml2ora as been renamed to compile (and has a new mode argument for specifying the database type).

• The targetroot parameter for ll.make.XISTConvertAction.__init__ has been renamed to root.

• Added the missing file _misc_include.c to the distribution archives.

• A new action class EvalAction has been added to ll.make.

• ll.xist.helpers.escapetext and ll.xist.helpers.escapeattr have been merged into one function that escapes all special characters (including ' and ") and has been renamed/moved to ll.misc.xmlescape.

• Python versions of all the functions in the module ll.misc have been added. Those versions will be used in case the C module is not available.

• A refcounting bug in the attribute parsing code of sgmlop has been fixed.

• The helper function cssescapereplace has been removed, as it's no longer needed.

• Pure Python versions of helpers.excapetext and helpers.escapeattr have been added, in case the C module is not available.

• The following functions have been added to ll.xist.css: parsestring, parsestream, parsefile, parseurl and write. They parse CSS resources and are similar to the XML/HTML parsing functions in that they apply the specified base URL to all URLs in the style sheet.

• cssutils 0.9.5b2 is required now.

• ll.xist.css.iterrules and ll.xist.css.applystylesheets now support specifying whether the preferred stylesheets or an alternate stylesheet group should be used.

• ll.xist.xsc.ProcInst.__mul__ and ll.xist.xsc.ProcInst.__rmul__ now return a fragment containing the node repeated a number of times instead of one processing instruction node containing repeated content.

• The constructor for ll.xist.parsers.ExpatParser now takes two additional arguments:

  xmldecl

  If this is true the XML declaration will appear in the resulting XIST tree.

  doctype

  If this is true the doctype declaration will appear in the resulting XIST tree (however any internal DTD subset will be dropped).

• cssutils 0.9.5 is used now. This simplifies the implementation of css.selector.

• A function ll.xist.css.geturls has been added. This returns a list of all the URLs in a cssutils stylesheet.

• toxic.xml2ora now treats unknown processing instructions as text. This makes it possible to e.g. output an XML header via toxic.

• The pseudo-elements in ll.xist.ns.jsp are no longer in a namespace, so they will always be published without any prefixes.

• A new method replaceurls has been added to ll.xist.xsc.StyleAttr. With this method all URLs in a style attribute can be replaced.

• Fixed a bug in ll.xist.parsers.SGMLOPParser.begin: The encoding wasn't passed properly to the XML decoder.

• ll.xist.xsc.ProcInst.publish now calls the checkvalid method too.

• It's now possible to force the publisher to output certain xmlns attributes via the showxmlns argument to the Publisher constructor.

• The core package has been moved into XIST, installing XIST now only requires one package.

• ll.toxic has been moved into XIST and is now available as ll.xist.ns.toxic.

• When a ll.make.XISTParseAction object is executed the content of the pool will now be extended by the content of the pool from the XISTPoolAction instead of being replaced.

• ll.make.Pool and ll.xist.xsc.Pool no longer use a WeakValueDictionary, but a simple dict. This means they can now store any object. A method clear has been added, which removes all registered objects.

• Fixed a bug in ll.xist.css.iterrules that surfaced when a base argument was given.

• Fixed a second bug in ll.xist.css.iterrules where the href of a link element wasn't applied to the URLs in the stylesheet.

• Fixed the problem that the source distibution didn't include header files.

• If an URLAttr attribute contains a processing instruction XIST will no longer transform the URL in any way.

• Fixed a parser bug where attributes were dropped when the attribute value was empty.

• Putting a module into a Pool object now copies the xmlns attribute too. This makes it possible to use Pool objects as conversion targets.

• Namespaces have been greatly simplified. There are no namespace modules any longer. An element class can be assigned a namespace by setting the xmlns class attribute to the namespace name. Global attributes can be assigned a namespace by setting the xmlns attribute on the attribute class itself (not on the Attrs class). The classes Prefixes and NSPool are gone too. Instead a new class Pool is used to specify which classes should be used for parsing.

• Dependency on PyXML has finally been dropped. XIST now uses its own XML parsing API. Two parsers are available: One based on expat and one based on a custom version of sgmlop.

• Tree traversal has been rewritten again. XFind expressions involving multiple uses of // now work correctly. The method walk now doesn't yield Cursor objects, but simple path lists (actually it's always the same list, if you want distinct lists use walkpath). Applying XFind expressions to nodes directly is no longer supported, you have to call walk, walknode or walkpath with the XFind expression instead. Many XFind operators have been renamed and/or reimplemented (see the documentation for the xfind module for more information).

• The methods __getitem__, __setitem__ and __delitem__ for Frag and Element now support the new walk filters, so you can do:

  • del node[html.p] to delete all html.p child elements of node;

  • del node[html.p[2]] to delete only the third html.p;

  • node[xfind.hasclass("note")] = html.p("There was a note here!") to replace several child nodes with a new one;

  • for c in node[xfind.empty]: print c.bytes() to print all empty (element) children of node;

  • del node[node[0]] to delete the first child node (which is silly, but illustrates that you can pass a node to get/replace/delete that node);

• A new module ll.xist.css has been added which contains CSS related functionality: The generator function iterrules can be passed an XIST tree and it will produce all CSS rules defined in any html.link or html.style elements or imported by them (via the CSS rule @import). This requires the cssutils package.

• The function applystylesheets modifies the XIST tree passed in by removing all CSS (from html.link and html.style elements and their @imported stylesheets) and putting the styles into style attributes of the affected elements instead.

• The function selector return a tree walk filter from a CSS selector passed as a string.

• Constructing trees can now be done with with blocks. Code looks like this:

  with xsc.Frag() as node:
      +xml.XML()
      +html.DocTypeXHTML10transitional()
      with html.html():
          with html.head():
              +meta.contenttype()
              +html.title("Example page")
          with html.body():
              +html.h1("Welcome to the example page")
              with html.p():
                  +xsc.Text("This example page has a link to the ")
                  +html.a("Python home page", href="http://www.python.org/")
                  +xsc.Text(".")
  
  print node.conv().bytes(encoding="us-ascii")

  Also the function xsc.append has been renamed to add and supports with blocks now instead of XPython.

• A subset of ReST is supported now for docstrings when using the ll.xist.ns.doc module. The module attribute __docformat__ is now honored (Set it to "xist" to get XIST docstrings).

• Many classes in the ll.xist.ns.doc have been renamed to more familiar names (from HTML, XHTML 2 or ReST).

• The media attribute of html.link and html.style now has a method hasmedia.

• The node method asBytes has been renamed to bytes and bytes has been renamed to iterbytes.

• The node method asString has been renamed to string and a new method iterstring has been added.

• ll.xist.ns.xml.XML10 is gone now. Use ll.xist.ns.xml.XML instead.

• xsc.tonode now will raise an exception when it can't handle an argument instead of issuing a warning.

• A class attribute empty inside element classes will now no longer get converted into model.

• ll.xist.ns.doc.pyref now copes better with decorated methods.

• The deprecated Element methods hasAttr, hasattr, isallowedattr, getAttr, getattr, setDefaultAttr, setdefaultattr, attrkeys, attrvalues, attritems, iterattrkeys, iterattrvalues, iterattritems, allowedattrkeys, allowedattrvalues, allowedattritems, iterallowedattrkeys, iterallowedattrvalues, iterallowedattritems and copyDefaultAttrs have been removed. The deprecated Attrs method copydefaults has been removed too.

• The namespace module ll.xist.ns.cond has been removed.

• When calling the function ll.xist.parsers.parseURL the arguments headers and data are now passed along to the parser's method only if they are specified. This makes it possible to pass ssh URLs to ll.xist.parsers.parseURL.

• The methods withnames and withoutnames have been split into two that take Python names and two that take XML names. Multiple arguments are used now (instead of one argument that must be a sequence). Passing a namespace to remove all attributes from the namespace is no longer supported.

• The Attrs methods updatenew and updatexisting have been removed.

• The Python quotes example no longer contains the XML source or the generated HTML.

• The Python quotes example now always parses the file from the original URL.

• The Python quotes and the media example now print the result to stdout.

• Use a consistent license (MIT) everywhere. This should make XIST Debian compatible.

• Change the Python quotes example, so that it works even if there's no python-quotes.xml in the current directory.

• Fixed a bug in presenters.CodePresenter.__str__.

• Fixed base URL handling for tidy parsing.

• Updated examples.

• Updated xiter and xattrs implementations for Node and Namespace to conform to the newest version of IPython.

• Fixed a few bugs in the sgmlop function declarations.

• Readded the spacer pixel.

• XIST has been made compatible with Python 2.5: Code has been updated to use the proper C API for memory management and PEP 353 support has been added. XIST now includes its own fixed version of sgmlop.

• The ll.xist.xsc.Attrs methods with and without have been renamed to withnames and withoutnames for Python 2.5 compatibility.

• ll.xist.ns.htmlspecials.pixel no longer handles colors via different GIFs. It uses the background-color in the style attribute instead. The same change has been implemented for ll.xist.ns.htmlspecials.autopixel. It's now possible to overwrite the default src attribute value of root:px/spc.gif either via the XML attribute or via the converter context.

• The node method asText has been made a function, moved into the html namespace and renamed to astext. Furthermore elinks is used for plain text formatting now instead of w3m.

• Fixed a bug in the presentAttr method of ll.xist.presenters.TreePresenter.

• Fixed a bug in the presentEntity method of ll.xist.presenters.CodePresenter.

• Updated installation instructions.

• Namespaces for RSS 0.91, RSS 2.0 and Atom 1.0 have been added.

• A new namespace ll.xist.ns.detox has been added that is similar to ll.toxic but can be used to generate Python code instead of PL/SQL code. Using detox templates is about 50 times faster than using XIST trees directly and about 10 times faster than Kid.

• Presenters are now compatible to IPython ipipe module. This means that you can browse XIST trees interactively if you have IPython installed. NormalPresenter and the Node methods repr and asrepr have been removed.

• A new processing instruction ll.xist.ns.specials.url has been added that does the same URL transformation as ll.xist.xsc.URLAttr does.

• On publishing ll.xist.ns.html.html now only adds a lang and xml:lang attribute, if neither of them exists.

• setuptools is now supported for installation.

• ll.xist.xsc.Namespace.tokenize requires a unicode object as input now. This makes it possible to use encodings that are not ASCII compatible (such as UTF-16). The encoding argument is gone.

• ll.xist.xsc.Node.asString uses the encoding argument to determine which characters have to be output as character references now. (You'll still get a unicode object as the result.)

• A new processing instruction class ll.xist.ns.specials.literal has been added, that will output its content literally when published. This can be used for embedding preformatted XML (e.g. from a database) into an XIST tree.

• Namespaces for Relax NG and Kid have been added.

• XIST requires version 1.0 of the core package now.

• The class name for the DocBook DTD class has been fixed.

• A script xml2xsc.py has been added, that can be used to parse an XML file and generate a rudimentary XIST namespace from it.

• A DocType for XHTML 1.1 has been added (suggested by Elvelind Grandin).

• Line number information is now added when parsing HTML.

• The sorted method now supports the same arguments (cmp, key and reverse) as list.sort and sorted in Python 2.4.

• The walk doesn't yield the node directly, but yields a Cursor object now, with has several ways of referencing the node.

• New methods walknode, walkpath and walkindex have been added.

• Presenters use an iterator API instead of a stream API now. Dumping an XML tree presentation to the terminal can now start immediately instead of having to wait for the complete string to be formatted.

• Fixed a bug with element/attribute names that contained a . character. (This broke ll.xist.ns.fo.)

• Fixed a bug with xmlns attributes in nested elements. When an element ended the parser restored the wrong prefix mapping.

• The python-quotes demo has been updated to use the current version of AMK's XML file.

• Removed iterator stuff from ll.xist.xfind, as this is now part of the ll package/module.

• The function ToNode has been renamed to tonode.

• ll.xist.Context no longer subclasses list.

• ll.xist.ns.doc.explain will now try to output the objects in the order in which they appear in the Python source.

• The node methods find and findfirst have been removed.

• ll.xist.ns.cond now uses a sandbox dictionary in a converter context for evaluating expression.

• The content of the processing instruction ll.xist.ns.code.pyexec will not be executed at construction time, but at conversion time. The code in ll.xist.ns.code.pyexec or ll.xist.ns.code.pyeval will no longer be executed in the ll.xist.sandbox module (which has been removed), but in a sandbox dictionary in the converter context of the ll.xist.ns.code namespace.

• The tests have been ported to py.test.

• The method mapped is now callable without arguments. In this case a converter will be created on the fly. You can pass constructor arguments for this converter to mapped as keyword arguments.

• The publishing API has changed again: ll.xist.publishers.Publisher.publish no longer accepts an argument stream to which the byte strings are written, but it is a generator now. The publisher methods write and writetext have been renamed to encode and encodetext and return the encoded byte string, instead of writing it directly to the stream. There's a new generator method bytes for nodes now, which can be passed the same arguments as asBytes. These changes should help when using XIST in WSGI applications.

• The iterator returned from Element.__getitem__, Frag.__getitem__ and the walk method now supports __getitem__ itself, so you can write table[html.tr][0] to get the first row from a table or page.walk(xsc.FindTypeAll(html.td))[-1] to get the last table cell from a complete HTML page.

• Several bugs in the namespaces ll.xist.ns.meta, ll.xist.ns.form and ll.xist.ns.specials have been fixed.

• The namespace modules ll.xist.ns.css and ll.xist.ns.cssspecials have been removed.

• XIST trees can now be pickled. The only restriction is that global attributes must come from a namespace that has been turned into a module via makemod, so that this module can be imported on unpickling.

• Two arguments of the walk method have been renamed: filtermode has been renamed to inmode and walkmode has been renamed to outmode. For these modes two new values are supported:

  ll.xist.xsc.walkindex

  The value passed to the filter function or yielded from the iterator is a list containing child indizes and attribute names that specify the path to the node in question.

  ll.xist.xsc.walkrootindex

  The filter function will be called with two arguments: The first is the root node of the tree (i.e. the node for which walk has been called), the second one is an index path (just like for ll.xist.xsc.walkindex). If used as an outmode a tuple with these two values will be yielded.

• Attribute mappings now support __getitem__, __setitem__ and

  __delitem__ with list arguments, i.e. you can do:

  >>> from ll.xist.ns import html
  >>> e = html.a("gurk", href=("hinz", "kunz"))
  >>> print e.attrs[["href", 0]]
  hinz
  >>> e.attrs[["href", 0]] = "hurz"
  >>> print e["href"]
  hurzkunz
  >>> del e.attrs[["href", 0]]
  >>> print e["href"]
  kunz

  XML attributes can now be accessed as Python attributes, i.e.:

  >>> from ll.xist.ns import html
  >>> e = html.a("spam", href="eggs")
  >>> print e.attrs.href
  eggs

  (Don't confuse this with e.Attrs.href which is the attribute class.)

• Frag and Element now support Node subclasses as arguments to their __getitem__ method: An iterator for all children of the specified type will be returned.

• The encoding used for parsing now defaults to None. When reading from an URL and no default encoding has been specified the one from the Content-Type header will be used. If this still doesn't result in a usable encoding, "utf-8" will be used when parsing XML and "iso-8859-1" will be used when parsing broken HTML.

• All error and warning classes from ll.xist.errors have been merged into ll.xist.xsc. This avoids import problems with circular imports.

• The attributes showLocation and showPath of ll.xist.presenters.TreePresenter have been lowercased and presenters are properly reset after they've done their job.

• The class attribute xmlname will no longer be turned into a list containing the Python and the XML name, but will be the XML name only. You can get the Python name of foo from foo.__class__.__name__.

• DeprecationWarnings for name and attrHandlers have finally been removed.

• Instances of ll.xist.xsc.Entity subclasses can now be compared. __eq__ simply checks if the objects are instances of the same class.

• Added a note about the package init file to the installation documentation.

• XIST requires Python 2.4 now.

• ll.xist.ns.specials.x has been renamed to ll.xist.ns.specials.ignore.

• ll.xist.utils.findAttr has been renamed to ll.xist.utils.findattr.

• ll.xist.xfind.item no longer handles slices.

• XFind has been enhanced to support item and slice operators, i.e. if foo is an XFind operator, foo[0] is an operator that will produce the first node from foo (if there is one). Negative values and slices are supported too.

• Operators can be chained via division: html.a/html.b is an operator that can be passed around and applied to a node.

• XIST requires the new core module and makes use of the new "cooperative displayhook" functionality defined there: If you install the displayhook you can tweak or replace ll.xist.presenters.hookpresenter to change the output.

• The transparent pixel used by ll.xist.ns.htmlspecials.pixel has been renamed to spc.gif to avoid problems with IE.

• Removed a debug print in ll.xist.xfind.Finder.__getitem__.

• ll.xist.xfind now has a new function item, that can be used to get a certain item or slice from an iterator. xfind.first and xfind.last have been changed to use xfind.item, so you now have to pass a default value to get the old behaviour.

• Obsolete options in ll.xist.options have been removed (and reprEncoding has been renamed to reprencoding).

• Fixed a bug in ll.xist.parsers.Parser.parse.

• Fixed a bug in ll.xist.xfind.Finder.__floordiv__.

• Restricted characters as defined in XML 1.1 will now be published as character references.

• ToNode now tries iterating through the value passed in, so it's now possible to pass iterators and generators (and generator expressions in Python 2.4) to Frag and Element constructors.

• A new API named XFind has been added for iterating through XML trees. XFind expressions look somewhat like XPath expressions but are pure Python expressions. For example finding all images inside links in an HTML page can be done like this:

  from ll.xist import parsers, xfind
  from ll.xist.ns import html
  node = parsers.parseURL("http://www.python.org/", tidy=True)
  for img in node//html.a/html.img:
      print img["src"]

• The module ll.xist.xfind contains several operators that can be used in XFind expressions.

• Parsing broken HTML is now done with the HTML parser from libxml2. The parsing functions no longer accept options for tidy, only the boolean value of the tidy argument is used.

• The publishing API has been simplified: Publication can now be done with a call to ll.xist.publishers.Publisher.publish, passing in a ll.xist.xsc.Node. Writing strings to the publisher output is now done with ll.xist.publishers.Publisher.write. The methods beginPublication and endPublication have been removed.

• The presentation API has been simplified in the same way: You'll get a presentation by calling: string = presenter.present(node). The methods beginPresentation and endPresentation have been removed.

• The parser now has the option to ignore illegal elements, attributes, processing instructions and entities. The default behaviour is to raise an exception, but this can now be reconfigured via Python's warning framework.

• The classmethod tokenize from ll.toxic has been moved to ll.xist.xsc.Namespace, so it's now possible to tokenize an XML string for other processing instructions as well.

• A new class ll.xist.xsc.NSPool has been added. An NSPool contains a pool of namespaces from which the parser selects the appropriate namespace once an xmlns attribute is encountered.</item>

• The script xscmake.py (which has been unmaintained for a while now) has been removed.</item>

• Elements hostname, tty, prompt and input were added to ll.xist.ns.doc.

• The method ll.xist.xsc.Attrs.set now returns the new attribute object.

• The visit method has been removed.

• ll.xist.xsc.FindOld has been removed.

• ll.xist.ns.xml.header has been renamed to ll.xist.ns.xml.declaration.

• Specifying content models for elements has seen major enhancements. The boolean class attribute empty has been replaced by an object model whose checkvalid method will be called for validating the element content.

• A new module ll.xist.sims has been added that provides a simple schema validation. Schema violations will be reported via Pythons warning framework.

• All namespace modules have been updated to use sims information. The SVG module has been updated to SVG 1.1. The docbook module has been updated to DocBook 4.3.

• It's possible to switch off validation during parsing and publishing.

• ll.xist.xsc.Frag and ll.xist.xsc.Element both have a __call__ method with the same arguments as their constructors. Those methods will append content nodes (and set attributes for ll.xist.xsc.Element) and return self, so they can be used when creating an object tree. This makes it possible to put the attributes close to the tag name, instead of putting them at the end after the content.

  Instead of:

  node = html.table(
      html.tr(
          html.td("foo"),
          html.td("bar"),
      ),
      html.tr(
          html.td("spam"),
          html.td("eggs")
      ),
      class_="example"

  you can now use the following:

  node = html.table(class_="example")(
      html.tr(
          html.td("foo"),
          html.td("bar"),
      ),
      html.tr(
          html.td("spam"),
          html.td("eggs")
      )
  )

• Experimental support for Holger Krekel's XPython has been added. Code might look like this:

  from ll.xist import xsc, converters
  from ll.xist.ns import html, meta
  
  import random
  
  c = converters.Converter()
  <c>:
      <html.html()>:
          <html.head()>:
              <meta.contenttype()>: pass
              <html.title()>:
                  xsc.append("The title")
          <html.body(class_="foo")>:
              <html.h1()>:
                  flag = random.choice((0, 1))
                  if flag:
                      xsc.append("The foo page", class_="foo")
                  else:
                      xsc.append("The bar page", class_="bar")
              <html.p()>:
                  if flag:
                      xsc.append("The foo content")
                  else:
                      xsc.append("The bar content")
  
  print c.lastnode.asBytes()

• Creating global attributes has been simplified. Passing an instance of ll.xist.xsc.Namespace.Attrs to an Element constructor now does the right thing:

  from ll.xist.ns import html, xml
  node = html.html(
      html.head(),
      xml.Attrs(lang="de"),
      lang="en",
  )

• Creating skeleton implementations of XIST namespaces is no longer done via XML conversion (i.e. the namespace module ll.xist.ns.xndl), but through the new module ll.xist.xnd. The script dtdxsc.py will automatically generate sims information.

• ll.xist.xsc.CharRef now inherits from ll.xist.xsc.Text too, so you don't have to special case CharRefs any more. When publishing, CharRefs will be handled like Text nodes.

• ll.xist.ns.meta.contenttype now has an attribute mimetype (defaulting to "text/html") for specifying the MIME type.

• ll.xist.ns.htmlspecials.caps has been removed.

• Registering elements in namespace classes has been rewritten to use a cache now.

• Pretty printing has been changed: Whitespace will only be added now if there are no text nodes in element content.

• Two mailing lists are now available: One for discussion about XIST and one for XIST announcements.

• Changed the xmlname of ll.xist.ns.jsp.directive_page back again (it's directive.page only for the XML form, which we don't use anyway.)

• Drop the default value for ll.xist.ns.jsp.directive_page.Attrs.language, as this attribute can only be used once.

• If an ll.xist.xsc.Prefixes object has a prefix mapping for a namespace it will return this prefix too, if asked for a prefix for a subclass of this namespace.

• The class ll.xist.parsers.Handler has been renamed to Parser and has been made reusable, i.e. it is possible to instantiate a parser once and use it multiple times for parsing. All the classes derived from xml.sax.xmlreader.InputSource have been dropped and the methods for parsing strings, URLs and files have been implemented as methods of the parser. Most of the arguments that had to be passed to the various parsing functions are passed to the parser constructor now. The basic parsing functionality is implemented by parsing streams instead of InputSource objects.

• Similar to the changes for parsing, publishers have been changed to be reusable and most arguments to the publishing functions are available as arguments to the publisher constructor.

• Now converter contexts are no longer bound to an element class, but to the context class defined by the element class, i.e. the attribute Context of the argument for Converter.__getitem__ will be used as the dictionary key. This makes it possible to use a class and it subclasses interchangeably (as long as the base class defines its own Context class and the subclasses don't overwrite it).

• Added a find functor FindTypeAllAttrs that searches content and attributes.

• Fixed the XML name for ll.xist.ns.jsp.directive_page.

• All character references in ll.xist.ns.ihtml that exist in ll.xist.ns.chars too have been removed.

• It's now possible to parse XML without generating location information for each node, by passing loc=False to the constructor of the Handler.

• The HTMLParser no longer complains about global attributes or xmlns.

• XIST now supports uTidylib in addition to mxTidy. uTidylib is found it is preferred over mxTidy.

• It's possible now to pass arguments to tidy simple by passing an argument dictionary for the tidy argument in the parsing functions.

• The methods parsed and checkvalid have been separated.

• ll.xist.ns.htmlspecials.pixel and ll.xist.ns.htmlspecials.autopixel now check whether their color attribute is ok.

• The base URL is now set correctly when parsing from an URL even if the original URL does a redirect. (This requires ll.url version 0.11.3).

• Namespace handling has been rewritten again, to be more standards compliant: Now there is no prefixes for entities and processing instructions any longer. Prefix mappings can be created much simpler, and they no longer contain any namespace stack for parsing, as this is now done by the parser itself. xsc.NamespaceAttrMixIn is gone too.

• The processing instructions exec_ and eval_ from ll.xist.ns.code have been renamed to pyexec and pyeval and import_ has been removed.

• CharRefs from ll.xist.ns.html have been moved to a new module named ll.xist.ns.chars.

• The method names beginPublication, endPublication and doPublication have been lowercased.

• Namespace handling has been completely rewritten. Namespaces are now classes derived from ll.xist.xsc.Namespace. Defining element classes can be done inside or outside the namespace class. If the element classes are defined outside the namespace class, they can be moved inside the namespace with a simple attribute assignment:

  class foo(xsc.Element):
      empty = False
  
  class xmlns(xsc.Namespace):
      xmlname = "foo"
      xmlurl = "http://www.foo.com/ns/foo"
  
  xmlns.foo = foo

• The methods elementkeys, iterelementkeys, elementvalues, iterelementvalues, elementitems and iterelementitems can be used for iterating through the element classes and their names. You can use the method element to get an element class with a certain name:

  >>> from ll.xist.ns import html
  >>> html.element("div")
  <element class ll.xist.ns.html/div at 0x824363c>

• For processing instructions, entities and character references similar methods are available.

• The method update can be used to add many element classes to a namespace at once, simply by passing a dictionary with those classes (use vars() to add everything that's defined inside your module). The method updatenew does the same, but copies only those attributes that don't exist in the namespace, updateexisting copies only those that do exist. You can turn a namespace into a module with makemod:

  from ll.xist import xsc
  
  class foo(xsc.Element):
      empty = False
  
  class xmlns(xsc.Namespace):
      xmlname = "foo"
      xmlurl = "http://www.foo.com/ns/foo"
  xmlns.makemod(vars())

• Put the above code into foo.py and you can do the following:

  >>> import foo
  >>> foo
  <namespace foo/xmlns name=u'foo' url=u'http://www.foo.com/ns/foo' with 1 elements from 'foo.py' at 0x81bfc14>

• getns has been dropped, so you always have to pass in a Namespace class where a namespace is required.

• For the ll.xist.ns.jsp.directive_page element automatic generation of the correct charset option in the contentType attribute is only done when there is a contentType attribute, as contentType is optional.

• The converter has a new property node. node can't be passed to conv but will be set to self by conv automatically. This makes it possible to access the "document root" during conversion.

• ll.xist.ns.htmlspecials.autoimg no longer touches existing width and height attributes. This means that %-formatting of the existing attributes is no longer done.

• Added a new class ll.xist.ns.htmlspecials.autopixel that works like ll.xist.ns.htmlspecials.pixel but inherits the size for the image specified via the src attribute.

• Frag and Element now support extended slices.

• Frag and Element now support the methods extend and __iadd__.

• For walking the tree the method walk has been completely rewritten and a new method visit has been added. For more info see the docstrings.

• Node now has two new methods copy and deepcopy and supports the copy module from the Python standard library.

• Calling mapped through conv has been removed. You again have to call mapped directly and pass a node and a converter.

• The HTML handling of the HTMLParser has been improved (it now uses code from xml.sax.drivers2.drv_sgmlop_html (which is part of PyXML.

• The core functionality found in the script dtd2xsc.py has been moved to a class method ll.xist.ns.xndl.fromdtd in the ll.xist.ns.xndl namespace.

• ll.xist.parsers.ExpatParser is now a real subclass instead of an alias for xml.sax.expatreader.ExpatParser It reports unknown entity references to the application (if loading of external entities is switched off, which is done by ll.xist.parsers.Handler and only outside of attributes).

• Namespaces have been added for Zope's TAL and METAL specifications.

• A namespace has been added for XSL-FO.

• Remove the checks for attributes in attributes and moved the publication code for the full element into a separate method. This allows JSP tag library namespaces to simply overwrite publish to publish the element even inside attributes. (This is the same fix as in release 1.5.10).

• The methods sorted, reversed and shuffled have been rewritten so they no longer use sys.maxint. This change fixes those methods for 64 bit platforms (reported by Giles Frances Hall)

• ll.xist.ns.struts_config11.plug_in now allows content (as the DTD states). (This is the same fix as in release 1.5.8.)

• Added a few elements and attributes to ll.xist.ns.doc: username, which is used for the name of a user account, xref, which is used for internal cross references and the attribute id for section, which specifies the target for an xref.

• Added a new namespace module ll.xist.ns.xndl that contains the "XIST namespace definition language", i.e. elements that describe an XIST namespace and can be used by various scripts to generate skeleton namespace modules. The first of these script is the DTD to namespace converter dtd2xsc.py.

• Added a new namespace module ll.xist.ns.tld that contains the definition for Java Server Pages Tag Library descriptors and a script tld2xsc.py that uses this namespace to generate namespace modules from tld files.

• Attr now supports the method filtered. This is used by without now. The arguments for without have changed, because handling global attributes was too "magic". A new method with has been added, with does the opposite of without, i.e. it removes all attributes that are not specified as parameters.

• The Python name of each Node subclass is now available as the class attribute pyname.

• To continue the great renaming withSep has been renamed to withsep.

• The namespace name for the ll.xist.ns.struts_html module has been fixed.

• The argument defaultEncoding for the various parsing functions has been renamed to encoding.

• ll.xist.ns.doc.getDoc has been renamed to getdoc.

• The CSS parser was dropping the % from percentage values. This has been fixed.

Changes in 2.0.7 (released 11/12/2002)

• xsc.Element.__nonzero__ can no longer fall back to xsc.Frag.__nonzero__. (this is the same fix as in 1.5.7).

• Performance optimizations.

• Fixed a bug in ll.xist.ns.specials.autoimg: Attributes were not converted before the size check was done (this is the same fix as in 1.5.5).

• Fixed a regression bug in ll.xist.ns.jsp.directive and several documentation issues.

• Fixed a few bugs in HTMLParser.

• Added DocBook conversion for several elements in ll.xist.ns.doc.

• Now the __init__.py file for the ll package is included.

• Fixed a bug in Frag.__rmul__ (by reusing __mul__).

• Fixed a bug with the backwards compatible prefix mapping: Defining element classes in exec processing instructions didn't work, because the prefixes object used for parsing wouldn't be updated when the namespace object is defined inside the processing instruction. Now using the default for the prefixes argument in calls to the parsing functions uses one global shared Prefixes instances where all the namespaces that are newly defined will be registered too.

• Fixed xscmake.py by removing the prefix handling. OldPrefixes will always be used for parsing now.

• XIST now requires at least Python 2.2.1.

• Attribute handling has been largely rewritten. Instead of a class attribute attrHandlers, the attributes are now defined through a nested class named Attrs inside the element. This class must be derived from ll.xist.Element.Attrs (or one of its subclasses if you want to inherit attributes from this class). Defining attributes is done through classes nested inside this attributes class and derived from any of the known attribute classes (like TextAttr, URLAttr etc.). The class name will be the attribute name (and can be overwritten with a class attribute xmlname. This makes it possible to have docstrings for attributes. Furthermore it's possible to define an attribute default value via the class attribute default, allowed values for the attribute via values, which is a list of allowed values, and whether the attribute is required or not via required.

• XIST now has real namespace support. The new class ll.xist.xsc.Prefixes can be used to define a mapping between prefixes and namespace names. This can be used for parsing and publishing. Namespace support is even available for entities and processing instruction.

• Global attributes are supported now. Namespace modules for the xml and xlink namespaces have been added (and ll.xist.xsc.XML was moved to ll.xist.ns.xml).

• A new namespace module for SVG 1.0 has been added: ll.xist.ns.svg.

• The HTML specific parts of ll.xist.ns.specials have been split off into a separate module ll.xist.ns.htmlspecials.

• Comparison of attributes with strings has been removed. You have to use __unicode__ or __str__ now before comparing.

• The HTMLParser now removes unknown attributes instead of complaining.

• There is a new parser class BadEntityParser, which is a SAX2 parser that recognizes the character entities defined in HTML and tries to pass on unknown or malformed entities to the handler literally.

• To give all nodes a chance to do something after they have been parsed (e.g. to prepend the base URL for URLAttr nodes), the parser now calls the method parsed immediately after node creation. This is used for the new class StyleAttr, which uses the CSSTokenizer, to prepend the base URL to all URLs found in a style attribute.

• The pixel images have been moved to the directory px to make image URLs shorter.

• Updated to work with newer versions of ll.ansistyle.

• Updated the namespaces ll.xist.ns.struts_html and ll.xist.ns.struts_config11 to the state of Struts 1.1 final.

• Removed the default value for the className attribute in ll.xist.ns.struts_config11.action.

• Added an attribute type to ll.xist.ns.struts_config11.action_mapping.

• Implemented ll.xist.xsc.Namespace.__eq__, so that replacing a namespace in the registry really works.

• Added an attribute target to ll.xist.ns.html.area.

• Fixed a bug in the new ll.xist.ns.jsp.

• Updated ll.xist.ns.jsp to JSP 1.2.

• Remove the checks for attributes in attributes and moved the publication code for the full element into a separate method. This allows JSP tag library namespaces to simply overwrite publish to publish the element even inside attributes.

• Reregistering a namespace now properly overwrites the old version in xsc.namespaceRegistry.

• ll.xist.ns.struts_config11.plug_in now allows content (as the DTD states).

• xsc.Element.__nonzero__ can no longer fall back to xsc.Frag.__nonzero__.

• Performance optimizations.

• Fixed a bug in ll.xist.ns.specials.autoimg: Attributes were not converted before the size check was done.

• xscmake.py now tries to strip off a trailing xsc from the filename before it falls back to the extension html (The builtin extension mapping is still tried first).

• Added new processing instruction class ll.xist.ns.php.expression that generates a PHP print statement from its content.

• Removed the value magic from ll.xist.ns.form.checkbox as this conflicted with dynamic value values.

• Comparison of attributes with strings has been removed. You have to use __unicode__ or __str__ instead.

• The HTMLParser now removes unknown attributes instead of complaining.

• There is a new parser class BadEntityParser, which is a SAX2 parser that recognizes the character entities defined in HTML and tries to pass on unknown or malformed entities to the handler literally.

• To give all nodes a chance to do something after they have been parsed (e.g. to prepend the base URL for URLAttr nodes), the parser now calls the method parsed() immediately after node creation. This is used for the new class StyleAttr, which uses the CSSTokenizer, to prepend the base url to all urls found in a style attribute.

• The HTMLParser now removes unknown attributes instead of complaining.

• There is a new parser class BadEntityParser, which is a SAX2 parser that recognizes the character entities defined in HTML and tries to pass on unknown or malformed entities to the handler literally.

• To give all nodes a chance to do something after they have been parsed (e.g. to prepend the base URL for URLAttr nodes), the parser now calls the method parsed immediately after node creation. This is used for the new class StyleAttr, which uses the CSSTokenizer, to prepend to base URL to all URLs found in a style attribute.

• New namespace module xist.ns.struts_config11 allows to parse and modify Struts configuration files conforming to the Struts 1.1 DTD.

• Updated xscmake.py to be compatible with the new url module.

• xist.ns.jsp.directive_page now automatically sets the contentType on publishing.

• Removed TidyURLInputSource. Now it's possible to pass a tidy flag to the remaining functions parseString, parseFile and parseURL to specify whether the source should be tidied.

• To prevent an element from being registered in a Namespace the class attribute register can be used now. This makes it possible to have a name for the element even when it's not registered.

• xist.ns.form elements now have all the attributes that the corresponding elements from xist.ns.html have.

• Removed the old xist.url from the Windows distribution.

• Reimplemented URL handling again. Now the new global module url is used for that.

• Added a method pretty to Node for generating a pretty printable version of the node.

• xsc.Node.name no longer is a class method, but a class attribute, that will be set at class instantiation time by the meta class.

• Ported to Python 2.2. Node is now derived from object, Frag from list and there's a new class Attrs which is derived from dict for the attribute mappings. All presenters have been adapted to work with Attrs. In addition to the usual dictionary methods and operators Attrs has a method without that returns a copy of the Attrs instance with some specified attributes removed.

• All the node classes now have a new method walk that generates all nodes in the tree using the new generator feature of Python 2.2.

• Also a new method walkPath has been added that works the same as walk but yields the complete path to each node as a list.

• Added a class block to xist.ns.jsp. The content of the block instance will simply be enclosed in a {} block. xist.ns.php got such a class too.

• Added a new module xist.ns.ihtml for i-mode HTML.

• Added new modules xist.ns.css and xist.ns.cssspecials for generating CSS.

• Now the various attributes of the Converter object are collected in a ConverterState object and it's possible to push and pop those states, i.e. it's now easy to temporarily modify a converter object during a convert call and revert back to a previous state afterwards.

• parseURL and parseTidyURL now have an additional parameter headers which is a list of string pairs specifying additional headers to be passed in with the request.

• parseString has an additional parameter systemId which will be the system id of the InputSource.

• The distribution now includes the makefile and the XML source files so now the distribution can rebuild ifself.

• Various other small bugfixes and enhancements.

• Added a new element contentscripttype to xist.ns.meta that generates a <meta http-equiv="Content-Script-Type" ...> element.

• xist.ns.doc.explain now generates anchor elements for the class, function and method description, so now the links on the XIST webpages work.

• Docstrings and documentation has been reworked. Now xist.ns.doc.pyref no longer implies a font change. Use the classes xist.ns.doc.module, xist.ns.doc.class, xist.ns.doc.method, xist.ns.doc.function and xist.ns.doc.arg to mark up your Python identifiers.

• Added the attributes type and key to xist.ns.struts_config.data_source.

• Added the deprecated attributes start to xist.ns.html.ol and value to xist.ns.html.li.

• Added missing asPlainString methods to Comment and DocType.

• xist.url.URL.fileSize and xist.url.URL.imageSize now use the warning framework to report errors.

• There is a new presenter named CodePresenter that dumps the tree as Python source code.

• The filenames of the pixel images used by xist.ns.specials.pixel have changed. These images are now included.

• URLs that are completely dynamic will now be left in peace when parsing or publishing.

• xist.ns.meta.keywords and xist.ns.meta.description no longer call asPlainString on their content. This makes it possible to e.g. generate the keywords via JSP:

  >>> from xist import parsers
  >>> from xist.ns import meta, jsp
  >>> s = '<keywords>' + \
  ...     '<?jsp:expression "foo"?>' + \
  ...     '</keywords>'
  >>> e = parsers.parseString(s)
  >>> print e.conv().asBytes()
  <meta name="keywords" content="<%= "foo" %>" />

• When an element occurs inside an attribute during publishing, there won't be an exception raised any more. Instead the content of the element will be published. This fixes problems with abbreviation entities inside attributes.

• xist.parsers.TidyURLInputSource now uses the new experimental eGenix mx Extension package, which includes a Python port of tidy.

• __repr__ now uses the new class presenters.PlainPresenter which gives a little more info than the default __repr__.

• URL handling has been changed again. Upto now, URLAttr had an additional instance attribute base, which was the "base" file/URL from which the attribute was parsed. Now the base URL will be directly incorporated into the URL. You can pass the base URL to all the parsing functions. Similar to that when publishing you can specify a base URL. All URLs in the tree will be output relative to this base URL. Joining URLs is now done via __div__ and no longer via __add__. This makes it more consistent with fileutils. The plan is to make URLs string like immutable objects and to merge them with fileutils.Filename.

• xist.ns.specials.php has been moved to its own module (xist.ns.php). This module provided additional convenience processing instructions (just like xist.ns.jsp does).

• The global namespace registry now keeps a sequential list of all registered namespaces, which will be used by the parser when searching for names. This gives a predictable search order even without using Namespaces and its pushNamespace method: modules imported last will be searched first.

• Processing instructions are now allowed inside attributes when publishing.

• xist.ns.docbooklite has been renamed to xist.ns.doc. It can now generate HTML and Docbook output and has improved a lot. The XIST web pages now use this for automatic documentation generation. The doc example has been removed.

• xist.url.URL now has a new method info that returns the headers for the file/URL.

• xist.url.URL now has a methods fileSize and imageSize too.

• xist.ns.jsp.directive_page now has new attribute session.

• __repr__ now uses the new class presenters.PlainPresenter which gives a little more info than the default __repr__.

• Small bugfix in presenters.strProcInst.

• Fixed xist.ns.struts_html.option to allow content.

• Sequences in constructor arguments for Frag and Element are again expanded and it's again possible to pass dictionaries in an Element constructor to specify attributes. As sequences are always unpacked, the method extend is gone. This works for append and insert too.

• Node and Frag implement __mul__ and __rmul__, so you can do stuff like:

  html.br()*5

  This returns a Frag with five times to same node.

• Arguments for the converter constructor can be passed to xist.xsc.Node.conv now, so it's possible to do stuff like this:

  from xist.ns import code
  print code.Eval("return converter.lang").conv(lang="en").asBytes()

  which will print en.

• The option XHTML for the publishers has been changed to lowercase.

• xist.ns.html.html will automatically generate a lang and xml:lang attribute when the converter has a language set.

• New module for WML 1.3.

• The publishing interface has changed internally and publishing should be faster now.

• Publishers now support a new parameter: usePrefix, which specifies if namespace prefixes should be output for the element names.

• Part of the implementation of the publishing stuff has been moved to C, so now you'll need a C compiler to install XIST.

• When publishing ", it will now only be replaced with " inside attributes.

• All the asHTML methods now have an additional argument converter. This makes it possible to implement different processing modes or stages for new elements. All currently implemented elements and entities ignore this argument, but pass it on in the call to their childrens' asHTML method. As the name asHTML no longer makes sense, asHTML has been renamed to convert.

• There is now a tool dtd2xsc.py in the scripts directory that creates a skeleton XIST module from a DTD (this requires xmlproc from the PyXML package).

• New preliminary module for DocBook 4.12. (Incomplete: convert methods and Unicode character entities are missing; any volunteers for implementing 375 classes?)

• New module ruby.py that implements the W3C Ruby draft.

• sql.py has been removed from XIST, but is available as a separate module.

• The parsing interface has been changed. Parsing is now done with the functions parseFile, parseString, parseURL and parseTidyURL in the module parsers. It's now possible to specify which parser should be used for parsing by passing a SAX2 parser instance to any of these functions. XIST now includes a rudimentary SAX2 driver for sgmlop and a rudimentary HTML parser that emits SAX2 events.

• The python-quotes example has been updated to work with expat.

• Added a new example: media.

• All abbreviation entities have been moved to a new module abbr.py.

• All the modules that provide new elements and entitites have been moved to a subpackage ns.

• Frag and Element now have new methods sorted, reversed, filtered and shuffled that return sorted, reversed, filtered and shuffled versions of the Frag/Element object.

• New namespace modules ns/jsp.py and ns/struts_html.py have been added that allow you to use JSP and Struts tags with XIST.

• A new method asText was added, that returns the node as a formatted plain ASCII text (this requires that w3m is installed.)

• make.py has been renamed to xscmake.py and moved to the scripts directory, it will be installed as a callable script with python setup.py install_scripts.

• xscmake.py has a new option --files/-f. The argument is a file containing a list of filenames (one name per line) that should be converted.

• xscmake.py has a new option --parser/-r for specifying which parser to use. Allowed values are sgmlop and expat.

• xscmake.py has a new option --namespace/-n that can be used for appending Namespace objects to the Namespaces object used by xscmake.py:

  xscmake.py -n html -n spam eggs.xsc

  With this call the parser will find element classes from the module with the prefix name spam before those from html and those before anything else.

• xist.url.URL no longer has an attribute ext. file and ext are merged.

• The special treatment of sequences as constructor arguments to Frag and Element has been removed, so XIST will no longer remove one level of nesting. If you still want that, use a * argument.

• Frag and Element now have a new method mapped, that recursively maps the nodes through a function. This is like convert but via an external function.

• Attribute handling has been improved thanks to a suggestion by Hartmut Goebel: Element.__getitem__ now always works as long as the attribute name is legal. If the attribute is not set, an empty attribute will be returned. All empty attributes will be considered as being not set and so hasAttr returns 0 for them, and publish doesn't publish them. This simplifies several very common cases:

  • Copying an attribute from one element to another works regardless of whether the attribute is set or not;

  • Testing for an attributes presence can now be done much simpler: if element["attrname"] instead of if element.hasAttr("attrname") (which still works, and should be a little faster);

  • When you construct an XIST tree and the presence or absence of an attribute is tied to a condition, you can construct the attribute in advance and use it afterwards in the tree construction:

    if condition:
        align = "right"
    else:
        align = None
    node = html.div("spam", align=align)

    So, when the condition is false, the node will not have the attribute align set.

• xist.ns.cond.If (and xist.ns.cond.ElIf) can now be used to test for attributes of the converter. I.e. it's possible to write the following XML:

  <if lang="en">Title
  <elif lang="de">Überschrift
  </if>

• URL handling has be completely changed and is much, much simpler now. There are no more path markers. To specify an URL that is relative to the current directory use the scheme root (e.g. root:main.css).

• Fixed a bug in the entity handling.

• Added a few deprecated elements and attributes to the html module.

• Improved the publishing of attributes. Now all attribute values will be published. For boolean attributes no value will be published for XHTML==0 and the attribute name will be used for XHTML==1 or XHTML==2.

• Element.compact now works (better) ;).

• Incorparated many bug fixes from Hartmut Goebel.

• Implemented xsc.Element.copyDefaultAttrs, which copies unset attributes over from a dictionary (simplifies implementing specials.plaintable and specials.plainbody).

• providers.Provider.pushNamespace now handles multiple arguments which may be Namespace objects or modules (in which case, module.namespace will be pushed).

• providers.Providers.popNamespace can now pop multiple namespaces at once.

• providers.TidyURIProvider now uses os.popen3 for piping the file through tidy, so now there will be no more temporary files. The call to tidy now includes options that hopefully make the output more suited to XIST.

• Incorparated a new url.py by Hartmut Goebel, that fixes many problem (e.g. optimizing http://server/foo/bar/../../baz.gif now works.)

• make.py includes a new option --path for adding directories to sys.path.

• Now uses sgmlop.XMLParser instead of sgmlop.SGMLParser, so case is preserved.

• Fixed another regression from the URL to string conversion change.

• Converting URLs to nodes is now done in ToNode, so URL objects can be used everywhere.

• Fixed a few bugs in Text._strtext and URLAttr._str.

• Now testing if characters can be encoded with the specified encoding is done directy. This means, that escaping unencodable characters now works even with exotic encodings (tested with JapaneseCodecs 1.0.1.

• The URLAttr constructor now can handle a single parameter of the type URL.

• The URL to string conversion function have changed: URL.asString returns the URL with path markers, URL.asPlainString returns the URL without path markers.

• Added the i18n attribute to the font element.

• Fixed the clashes between the class names for the elements and entities sub and sup in html.py.

• Several small enhancements and bug fixes contributed by Hartmut Goebel.

• Now processing instruction classes are registered in the same way as elements and entities are.

• The leaf nodes (Text, Comment, ProcInst) are now considered immutable. This means that their asHTML method can simply return self, because now those nodes can be shared between trees. Functionality for manipulation the objects is provided by a mixin class very similar to UserString. All this results in a speedup of about 10% for the python-quotes example.

• Small optimizations in the asHTML methods of Element and Frag optimized away many calls to append, extend and ToNode and result in a speedup of about 30% for the python-quotes example. One consequence of this is that Null objects will no longer be ignored.

• New elements keywords and description in meta.py.

• Fixed a bug in Namespace.register, now setting name=None to prevent an element from being registered works again.

• A new module named meta.py has been created, that simplifies generating meta tags.

• Various small bugfixes.

• XIST now requires at least Python 2.0b1.

• A new bugfixed version of the sgmlop source is available from the FTP site.

• XIST now completely supports Unicode. For output any encoding known to Python can be used, so now you can output your HTML in ASCII, Latin-1, UTF-8, UTF-16, ...

• All publishers have been updated to support Unicode. The publishing interface has been streamlined (encoding and XHTML parameters are now attributes of the publisher).

• asString will now always return a Unicode string. If you want a byte string use asBytes instead, where the encoding can be specified as an argument.

• There an additional publisher class FilePublisher, which can be used for publishing to a file (or anything else that has a write and a writelines method, and is supported by the stream writer available through codecs.lookup).

• Element and attribute names are no longer converted to lowercase. If you have an attribute name which clashes with a Python keyword (e.g. class) append an underscore (_), which will be removed before accessing the attribute. This is the "official" Python method for handling these cases.

• Elements and entities are no longer registered one by one. Now you can build Namespace objects which are used for searching and there are pushNamespace and popNamespace functions in XSC.xsc. For more info, see the source.

• Image size calculation has been removed from html.img and html.input. Use specials.autoimg and specials.autoinput for that.

• __getitem__, __setitem__ and __delitem of Frag and Element now accepts a list as an argument. The method will be applied recursively, i.e. e[[0, 1, "foo", 2] is the same as e[0][1]["foo"][2].

• The deprecated module db.py no longer exists. Useful functions and elements from db.py have been moved to sql.py and form.py respectively.

• When using xsc.make the encoding and XHTML parameters to use can now be specified on the command line (e.g. --encoding utf-8 --xhtml 2)

• Handling of multiline <?xsc-eval?> and <?xsc-exec?> has been enhanced, although XIST will not be able to guess the correct indentation in all cases. As a workarround simply add a Python comment to the beginning. So the following won't work:

  <?xsc-exec
      for i in xrange(10):
          do(i)
  ?>

  But this will:

  <?xsc-exec
      #
      for i in xrange(10):
          do(i)
  ?>

• Make functionality has been moved to make.py, as certain modules can't be used as the main script, because reimporting them in processing instructions won't work. Now you can simply call:

  make.py --import xist.html --import spam eggs.xsc

• There is a new module cond.py, that contains elements that can be used for conditionals:

  <?xsc-exec a=42?>
  <if cond="a==21">
      <b>foo</b>
  <elif cond="a==42"/>
      <i>bar</i>
  <else/>
      baz
  </if>

• sgmlop will now be found either via import sgmlop or via from xml.parsers import sgmlop.

• Fixed a bug in URLAttr.publish, which prevented URLAttr from working at all.

• Fixed a bug in html.img and html.input. Now image size calculation works again.

• Changes in 0.3.6 (released 07/04/2000)

• Fixed a bug in Node._matches, which resulted in a non working find.

• The documentation example has been enhanced. Now documenting methods works.

• When the member elementname: in the element class is set before calling registerElement, this element name will be used for the element. This allows custom names even when using registerAllElements.

• Comparison of scheme and server in URLs is done case insensitive (as RFC 2068 requires.)

• Image size calculation is now done in asString and not in asHTML. This allows to write faster code. Old method:

  e = html.div(html.img(...),gurk.hurz()).asHTML().asString()

  New method:

  e = html.div(html.img(...),gurk.hurz().asHTML()).asString()

• Image size calculation is now done for <nput type="image">. The size attribute is set to the image width.

• Manipulating the path in an URL is now done via the usual __setitem__/__getitem__ stuff, which keeps the path in a consistent state:

  >>> from xist.URL import URL
  >>> u = URL("/foo/*/../bar/baz.gif")
  >>> del u[1]
  >>> u
  URL(scheme='server', path=['bar'], file='baz', ext='gif')

• findNodes (which has been shortened to find) has an additional argument test, which can be a test function that will be called when the node passes all other tests.

• asString no longer generates a string directly, but uses the new method publish, which has an additional argument publisher, to which the strings to be output are passed.

• Location information is now copied over in clone, asHTML and compact where appropriate, so you know even in the HTML tree where something came from.

• xsc.repransi can now have three values:

  0

  coloring is off

  1

  coloring is on for a dark background

  2

  coloring is on for a light background

• All repransi variables are now arrays with two strings, the first for dark, the second for light.

• The workaround for the trailing CDATA bug in sgmlop has been removed, so now you'll need a newer version of sgmlop (included in PyXML 0.5.5.1).

• These changes predate written history.